Sign in

Robert Nowak

@rdnowak.bsky.social
788 followers 111 following 47 posts

Director of the Center for the Advancement of Progress

PostsRepliesMedia
Robert Nowak @rdnowak.bsky.social · 21/04/2026
You should get the daily delivery of the Wisco state journal. It’s a great way to stay on top of local affairs
110
Robert Nowak @rdnowak.bsky.social · 15/02/2026
upload.wikimedia.org/wikipedia/en...
030
Robert Nowak @rdnowak.bsky.social · 29/12/2025
But, as I’ve heard from others, the AIs often suggest possible connections, related results, or avenues to pursue that hadn’t occured to me. Unfortunately, these are usually dead ends.
110
Robert Nowak @rdnowak.bsky.social · 29/12/2025
I’ve found that the pro-version AIs are great for proving known theorems or new theorems that could be considered homework problems, but so far I have had no success using them to solve truly open/novel/challenging math problems.
120
Robert Nowak @rdnowak.bsky.social · 17/12/2025
media.tenor.com
a man playing a guitar with the words heavy metal intensifies
ALT: a man playing a guitar with the words heavy metal intensifies
110
Reposted by Robert Nowak
Ben Recht @beenwrekt.bsky.social · 15/12/2025
Since I can’t get it out of my head, I wrote up my thoughts on @kevinbaker.bsky.social's critique of AI-automated science and the logical end of processes that can't self-correct.
argmin.net
Measures as ends
Some initial thoughts on Kevin Baker's provocative "Context Windows."
3439
Robert Nowak @rdnowak.bsky.social · 15/12/2025
Kevin Baker's essay is probably the best thing I have read in 2025.
010
Robert Nowak @rdnowak.bsky.social · 31/10/2025
Yes. Just write your thoughts in a rough and unpolished form, say rough paragraphs that contain terse points you want to make. then let 'er rip
110
Robert Nowak @rdnowak.bsky.social · 25/10/2025
Section 7 is a wonderful description of the process they went through.
110
Robert Nowak @rdnowak.bsky.social · 13/10/2025
something just isn't fully clicking. if you look at total yards and time of possession, they should have blown them out. well, better anyway to peak later in season, so let's hope that's what happens (like two seasons ago)
010
Robert Nowak @rdnowak.bsky.social · 13/10/2025
Packers get the win, but it wasn't pretty.
100
Robert Nowak @rdnowak.bsky.social · 08/09/2025
Thanks for participating and presenting your work!
010
Robert Nowak @rdnowak.bsky.social · 05/09/2025
Google promotes box shirts too
010
Robert Nowak @rdnowak.bsky.social · 27/08/2025
Pour into
020
Reposted by Robert Nowak
Dylan Foster 🐢 @djfoster.bsky.social · 11/08/2025
Announcing the first workshop on Foundations of Language Model Reasoning (FoRLM) at NeurIPS 2025! 📝Soliciting abstracts that advance foundational understanding of reasoning in language models, from theoretical analyses to rigorous empirical studies. 📆 Deadline: Sept 3, 2025
1103
Robert Nowak @rdnowak.bsky.social · 04/08/2025
Nice article about my mom’s new book shepherdexpress.com/culture/book...
shepherdexpress.com
Birdscaping for Wisconsin and the Great Lakes Region by Mariette Nowak
If you want the birds to flock to your garden, consider “birdscaping,” advises former director of Milwaukee’s Wehr Nature Center.
030
Robert Nowak @rdnowak.bsky.social · 05/07/2025
“the only way to predict or to control the functioning of such systems is by an intricate system of charms, spells, and incantations”
020
Robert Nowak @rdnowak.bsky.social · 21/06/2025
See you there!
020
Robert Nowak @rdnowak.bsky.social · 16/05/2025
More likely midges. The truest sign of a healthy ecosystem
110
Robert Nowak @rdnowak.bsky.social · 25/04/2025
Looking forward to a great MMLS!
030
Robert Nowak @rdnowak.bsky.social · 07/02/2025
This is collaboration with Ziyue Luo, @shroffness and @kevinlauka
000
Robert Nowak @rdnowak.bsky.social · 07/02/2025
Jifan’s on the industry job market now, and his expertise in efficient training, distillation, and data curation couldn't be more timely. Feel free to reach out to him at jifan@cs.wisc.edu. 📄 Paper: arxiv.org/abs/2410.02755
arxiv.org
GPT-4o as the Gold Standard: A Scalable and General Purpose Approach to Filter Language Model Pretraining Data
Large language models require vast amounts of high-quality training data, but effective filtering of web-scale datasets remains a significant challenge. This paper demonstrates that GPT-4o is remarkab...
110
Robert Nowak @rdnowak.bsky.social · 07/02/2025
SIEVE improves upon existing quality filtering methods in the DataComp-LM challenge, producing better LLM pretraining data that led to improved model performance. This work is part of Jifan's broader research on efficient ML training, from active learning to label-efficient SFT for LLMs.
100
Robert Nowak @rdnowak.bsky.social · 07/02/2025
Why does this matter? High-quality data is the bedrock of LLM training. SIEVE enables filtering trillions of web data for specific domains like medical/legal text with customizable natural language prompts.
100
Robert Nowak @rdnowak.bsky.social · 07/02/2025
SIEVE distills GPT-4's data filtering capabilities into lightweight models at <1% of the cost. Not just minor improvements - we're talking 500x more efficient filtering operations.
100
Robert Nowak @rdnowak.bsky.social · 07/02/2025
🧵 Heard all the buzz around distilling from OpenAI models? Check out @jifanz's latest work SIEVE - showing how strategic distillation can make LLM development radically more cost-effective while matching quality.
140
Robert Nowak @rdnowak.bsky.social · 23/01/2025
Maybe Trump should have read my mom's book: "For the first six weeks, the embryo, whether XX or XY, coasts along in sexual ambiguity." p. 25
020
Robert Nowak @rdnowak.bsky.social · 23/01/2025
www.rollingstone.com/politics/pol...
rollingstone.com
Trump's Trans Ban Defines Everyone as Female -- But That's Not the Problem
Donald Trump's executive order attempts to get rid of trans rights. LGBTQ+ advocates intend to fight back.
300
Reposted by Robert Nowak
Dimitris Papailiopoulos @dimitrisp.bsky.social · 18/01/2025
Task vectors are akin to punchcards: you feed them to your LLM and it implements specific tasks, without in-context demonstrations. Liu's new paper examines at what scale, where in the network and when during training do they emerge, and how to encourage their emergence. arxiv.org/pdf/2501.09240
0233
Robert Nowak @rdnowak.bsky.social · 04/01/2025
Good luck with that
100
Robert Nowak @rdnowak.bsky.social · 04/01/2025
p.s. we don't know for sure if I said this or not
110
Robert Nowak @rdnowak.bsky.social · 04/01/2025
Is the solution treating everything electronic as "fake"? Maybe?
110
Robert Nowak @rdnowak.bsky.social · 31/12/2024
I think the necessary length of the prompt(s) might also heavily depend on the data the model is trained on.
000
Robert Nowak @rdnowak.bsky.social · 28/12/2024
130
Robert Nowak @rdnowak.bsky.social · 28/12/2024
130
Robert Nowak @rdnowak.bsky.social · 24/12/2024
And it must have been addressed to prof Malik :)
010
Reposted by Robert Nowak
kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 21/12/2024
feeling a but under the weather this week … thus an increased level of activity on social media and blog: kyunghyuncho.me/i-sensed-anx...
kyunghyuncho.me
i sensed anxiety and frustration at NeurIPS’24 – Kyunghyun Cho
1917836
Reposted by Robert Nowak
Francis Bach @bachfrancis.bsky.social · 21/12/2024
A happy author discovering the first hard copies
39616
Reposted by Robert Nowak
Kaiping Chen @kaipingchen.bsky.social · 19/12/2024
🚨Excited to share our new publication @jmcquarterly.bsky.social. In our experiment, we showed promising evidence of how to design intergroup framing in news coverage to reduce conspiracy beliefs in the context of Sino-U.S. relationship👇 #journalism #scicomm journals.sagepub.com/doi/10.1177/...
173
Robert Nowak @rdnowak.bsky.social · 19/12/2024
Time for a meta office near the other UW :)
000
Reposted by Robert Nowak
Csaba Szepesvari @skiandsolve.bsky.social · 19/12/2024
If you are into ML theory (RL or not) with a proven track record, and you are interested in an industry research position, PM me. Feel free to spread the word.
27531
Robert Nowak @rdnowak.bsky.social · 11/12/2024
I’ll be presenting a poster on some work on OOD detection using active learning today at 4:30pst!
072
Reposted by Robert Nowak
Jifan Zhang @jifan-zhang.bsky.social · 11/12/2024
Check out our poster about humor and LLMs at #2908 today in the noon poster session! I’ll have to miss it, but Rob and Lalit will both be there to chat.
011
Robert Nowak @rdnowak.bsky.social · 10/12/2024
Since you are interested in discussing I guess it’s worth a read
000
Reposted by Robert Nowak
Data Science Institute @dsi-uchicago.bsky.social · 10/12/2024
We’ve extended the deadline for nominations to the Margot and Tom Pritzker Prize for AI in Science Research Excellence to January 5th, 2025! This prize celebrates exceptional contributions to AI and the natural sciences or engineering datascience.uchicago.edu/research/ai-...
083
Robert Nowak @rdnowak.bsky.social · 07/12/2024
Our NeurIPS paper studies differences in solutions from single-task/output neural network training vs. multi-task/output training. openreview.net/pdf?id=APBq3..., arxiv.org/pdf/2410.21696. Poster presentation time and day: Fri 13 Dec ,11am, East Exhibit Hall A-C #2303
openreview.net
080
Robert Nowak @rdnowak.bsky.social · 07/12/2024
Past work has characterized the functions learned by neural networks: arxiv.org/pdf/1910.01635, arxiv.org/abs/1902.05040, arxiv.org/abs/2109.12960, arxiv.org/abs/2105.03361. But it turns out multi-task training produces strikingly different solutions! Adding tasks produces “kernel-like” solutions.
17411
Robert Nowak @rdnowak.bsky.social · 04/12/2024
Is it safe to conclude that scaling suffices to make “fire” posters?
010
Robert Nowak @rdnowak.bsky.social · 04/12/2024
There is a ton of interest in the question of whether AI can be be funny: www.bbc.com/future/artic.... Our paper at NeurIPS investigates the humor generation capabilities of the latest and greatest AI models using one of world’s largest humor datasets! arxiv.org/pdf/2406.10522
091
Reposted by Robert Nowak
Yinglun Zhu @yinglunzhu.bsky.social · 03/12/2024
I’m recruiting multiple PhD students for Fall 2025 at UCR! If you’re interested in working on efficient ML, RL, and LLMs, please apply to the UCR CS/EE PhD program. Please visit yinglunz.com for detailed information on research directions and contact instructions.
032